This repo contains project prepared for the Bayesian Econometrics master classes at SGH Warsaw School of Economics (link). The classes were orginally tought using R and RStan but I was allowed to prepare the 2nd assignment in Julia and Turing.jl. I have choosen Julia as it provides rapidly developing packages ecosystem for bayesian modelling and enables preparing complete solution in one language, without splitting it into i.e. R and Stan. Julia also provides prospectous reactive notebooks engine, Pluto.jl which I have used for the purpose of this analysis.
This particular analysis focuses on providing an bayesian regression model with specified a priori expecations which explains differences in the irish whiskey sales dynamics across the countries. I used the data provided by the Irish Food Board in 2018, which consist of sales records from 1995 to 2017 for multiple countries and 3 quality categories. In this particular research I focused on EU14 (before Brexit) countries plus US and Canda.
For pre-rendered static .html reports (in Polish) check:
For source Pluto.jl notebook check:
TODO's:
- english translation of the report,
- forecasts and out-of-sample evaluation,
- additional analysis using Variational Inference.